Fix/llms (#366)

* Fix LMStudio, cache documents with tg-load-sample-documents

* Fix Mistral

trustgraph-flow/trustgraph/model/text_completion/cohere/llm.py

@@ -70,6 +70,8 @@ class Processor(LlmService):
                 model = self.model
             )
 
+            return resp
+
         # FIXME: Wrong exception, don't know what this LLM throws
         # for a rate limit
         except cohere.TooManyRequestsError:

trustgraph-flow/trustgraph/model/text_completion/lmstudio/llm.py

@@ -12,7 +12,6 @@ from .... base import LlmService, LlmResult
 
 default_ident = "text-completion"
 
-default_subscriber = module
 default_model = 'gemma3:9b'
 default_url = os.getenv("LMSTUDIO_URL", "http://localhost:1234/")
 default_temperature = 0.0

trustgraph-flow/trustgraph/model/text_completion/mistral/llm.py

@@ -86,6 +86,8 @@ class Processor(LlmService):
                 model = self.model
             )
 
+            return resp
+
         # FIXME: Wrong exception.  The MistralAI library has retry logic
         # so retry-able errors are retried transparently.  It means we
         # don't get rate limit events.